Unified Trade Up Analysis: 2017/2018

Richard Harries

Monday, 18 March 2019

Exploratory Data Analysis

This analysis combines data from the 29 community businesses that took part in Trade Up in 2017 with the 99 businesses that took part in 2018.

Trading ratios

Performance

Mean trading ratios fell for organisations in the Control Group and rose for those in the Match Trading Group. This is consistent with the core hypothesis of Trade Up.

Group n Baseline year Trade Up year Change
Control 30 57.5 54.6 -1.2
Match Trading 98 54.0 66.6 9.6

The same pattern occurs for the median trading ratios although the difference is slightly less marked, reflecting the fact that medians are less influenced by outliers (see boxplot below).

Group n Baseline year Trade Up year Change
Control 30 64.1 51.4 -2.8
Match Trading 98 51.7 71.9 6.9

Significance

Tests for significance can be unreliable for small samples. Two tests that are appropriate for the Trade Up dataset are the Kruskal-Wallis Test and the Mann-Whitney-Wilcoxon Test. Both tests lend some support to the core hypothesis of Trade Up. (Given the observed differences between the two groups, there is 10% chance that they both come from the same population - that is, that there is no actual difference between the groups.)

## 
##  Kruskal-Wallis rank sum test
## 
## data:  Trading Ratio Change by Group
## Kruskal-Wallis chi-squared = 2.783, df = 1, p-value = 0.09527
## 
##  Wilcoxon rank sum test with continuity correction
## 
## data:  Trading Ratio Change by Group
## W = 581, p-value = 0.09621
## alternative hypothesis: true location shift is not equal to 0

Plots

Distributions

Five community businesses in the Match Trading Group saw their trading ratios increase by more than 50% in 2018, compared to maximum growth of 40% in the Control Group. On the other hand, the largest trading ratio reduction also occured in the Match Trading Group, falling by more than 50%.

Trends

Median trading ratios went up in both the Control Group and the Match Trading Group across 2018, ending up between 75% and 80%. However, compared to the baseline year, quarterly trading ratios generally fell in both groups, with the Control Group doing relatively worse than the Match Trading Group. While this is not inconsistent with the core Trade Up hypothesis, it suggests that trading ratio growth might have a strong seasonal element across the four quarters observed.

Patterns

This final chart compares the increase/decrease in trading ratios with the “Eligible To Claim Ratio” (ie. the quarterly Trade Up grant as a proportion of total quarterly income). There should be no relationshhip between the two.